In recent years, in monitoring control systems (for example, SCADA: supervisory control and data acquisition), attention has been given to multi hop networks in which multiple node devices configured in a tree network topology are connected with each other in wireless ad hoc communication networks by using a carrier sense method. In a multi hop network, when the reliability of packet delivery to an adjacent node device is important, a routing algorithm is used in which a packet is transferred while acknowledging the delivery of the packet between the adjacent node devices.
With the routing algorithm in which data is transferred while acknowledging the delivery of the packet between the adjacent node devices, it is known that congestion occurs along the communication path of the packet. Consequently, in a wireless ad hoc network that communicates using the carrier sense method, because node devices are added in an autonomous distributed manner, the number of node devices increases. When the number of node devices increases, the number of requests to send a packet from one node device to a node device that corresponds to the top device of the tree in a tree network (hereinafter, referred to as a gateway (GW) device) increases. Then, requests to send a packet to the gateway device conflicts with requests to send a packet (ACK: acknowledgement packet) that is used for a packet delivery acknowledgement sent from the gateway device to peripheral node devices; therefore, the opportunity to transmit an ACK from the gateway device decreases. Consequently, a delay in ACK delivery from the gateway device increases and, if the delay time exceeds the ACK waiting time, the number of packets duplicated by peripheral node devices increases. If the number of packets duplicated increases, requests to send the duplicated packets further arise. Consequently, any delay of an ACK from the gateway device further increases and thus the generation of the duplicated packets become significant, which becomes a vicious circle.
In such a case, a monitoring control system may remain unusable for a long time until congestion is relieved. Accordingly, in order to stably operate the monitoring control system, it is important to suppress the occurrence of congestion and, if congestion occurs, it is important to set up a system that promptly ceases the congestion.
To cope with this, with conventional known technology, when the traffic volume of a wireless network is measured and the measured traffic volume exceeds a certain level, the transmission rate of packets sent from node devices is decreased.    Patent Document 1: Japanese National Publication of International Patent Application No. 2005-510956
However, in the conventional technology, an efficient way of suppressing the occurrence of packet congestion in wireless ad hoc networks is not considered.
Specifically, with the conventional technology, if a node device detects packet congestion in a network, the occurrence of congestion is suppressed by reducing the transmission rate of packets that are transmitted from the node device. With the conventional technology, if packet congestion in a network is detected, it is not possible to suppress the transmission rate of packets transmitted from peripheral node devices.
In contrast, in a tree network in which a gateway device is used as the top device of the tree, because packets transmitted from multiple node devices are concentrated at the gateway device, congestion tends to occur particularly in packet communication paths at the periphery of the gateway devices. At this point, as with the conventional technology, even if each of the node devices arranged at the periphery of the gateway device decreases the transmission rates of packets from their own node devices, this technology does not decrease the transmission rates by taking into consideration the congestion state of the packets concentrated at the gateway device. Accordingly, the congestion in the packet communication paths at the periphery of the gateway devices may sometimes not be efficiently suppressed. Furthermore, if an attempt to detect congestion is made at only one place, i.e., the gateway device, in accordance with the conventional technology, only the congestion occurring at the periphery of the gateway device is detected. On the other hand, if an attempt to detect congestion is made at each node device in accordance with the conventional technology, any congestion occurring in a network can be detected; however, in this case, a congestion detection function is needed for each of the node devices, which is inefficient in terms of the cost performance.